In vitro fertilization (IVF) carries an increased risk of multiple gestation; more than one half of the infants resulting from assisted reproductive technology are part of a multiple birth.1 Births conceived by IVF represent less than 1% of total births nationally, yet account for one third of twins and nearly 40% of triplets or higher order births.2 Multiple gestations are at increased risk for poor clinical outcomes, both maternal and neonatal.3 Furthermore, multiple births generate substantial health care costs. The average hospital cost for a triplet birth is more than ten times that of the average singleton birth.4
Increasing attention is being paid to lowering the IVF-associated multiple birth rate. Many countries, such as the United Kingdom, have legally mandated a maximal number of embryos that can be transferred per cycle. This has not been done in the United States, although the American Society for Reproductive Medicine has issued recommendations aimed at lowering the embryo transfer number.5 Although there has been a slight decline in the number of embryos transferred per cycle in the United States during recent years, multiple birth rates are still high, and the twin rate continues to rise.5 Thus, although researchers, policy makers, and clinicians are recognizing the need to lower multiple birth rates from IVF, there is a disconnection between this knowledge and its translation into clinical practice.
One reason for the difficulty in lowering the multiple birth rate is the presence of conflicting goals and incentives for the various parties involved. Several studies have shown that from a societal point of view, transferring one embryo at a time is more cost-effective than two or more.6–10 However, the infertile couple and their physician are actually making the decision. The infertile couple faces the high cost of IVF treatment, which is generally not covered by insurance, but does not face the high cost of multiple births, the bulk of which is covered by insurance. Furthermore, couples with infertility are likely focused on the short-term outcome of achieving pregnancy and are little concerned with the worse outcomes of multiple gestation.11 Interestingly, because infertility specialists are commonly compared by pregnancy rate per cycle, the physician may have an incentive toward increased number of embryos per cycle.
As we attempt to lower the multiple birth rate with a policy-driven approach, it is important to know not only the optimal embryo transfer decision from a societal point of view, but to see if it is in line with the incentives for insurance companies and parents, and if not, by how much. Recognizing and addressing possible conflicts of interest are important steps in formulating sensible health policy. Thus, our study addresses this question by analyzing the cost-effective number of embryos to transfer from various perspectives.
MATERIALS AND METHODS
A decision analytic model (Fig. 1) was developed with TreeAgePro 2004 software (TreeAge Software, Williamstown, MA). We used a Markov, or recursive model. The sequence of the model began with a woman undergoing one cycle of fresh embryo transfer with one to five embryos. If she failed to get pregnant or miscarried she underwent a second cycle, with the same number of embryos. The second cycle could be fresh or frozen, depending on whether she had remaining eggs from the first cycle. If she failed to get pregnant or miscarried she underwent a third cycle, which was fresh for all women. If she failed to get pregnant or miscarried she underwent a fourth cycle, which could be fresh or frozen. And finally, if she still failed to get pregnant, or miscarried, she underwent a fifth fresh cycle.
We created five different models, each with the same template but populated with different probabilities, for five maternal age groups: younger than 35, 35–37, 38–40, 41–42, and older than 42 years. These are the age categories provided in the Assisted Reproductive Technology Report.12 Our study was approved by the University of California, San Francisco Committee on Human Research.
In 1992, the Fertility Success Rate and Certification Act mandated that all fertility clinics in the United States report success rate data to the Centers for Disease Control and Prevention (CDC), which publishes an annual report. Our probability inputs are derived from nondonor cycles (for which complete information is provided) in the 2003 report,12 the most recent year for which data are available, and from the supplemental surveillance report from 2002,13 the most recent year that that is available (Table 1).
We derived the pregnancy rate from the live birth rate and miscarriage rate. Miscarriage rates were reported for every maternal age; we used the average rate for each maternal age subgroup. We made adjustments to the live birth rate data to better model our outcomes. Because the live birth rates were derived from cross-sectional data, it reflects, in part, that women with poorer prognosis or lower quality embryos had a greater number of embryos transferred. We, however, were modeling the decision for one woman. As such, the live birth rate should never be lower with an increased number of embryos transferred. Thus, we held the live birth rate constant for increasing embryo transfer number if the national data showed a decline.
Additionally, we adjusted the one-embryo transfer rate. For all age groups, the live birth rate for one-embryo transfers was less than one half that of two-embryo transfers. This is because the cross-sectional data reflects the fact that many of the women who had only one embryo transferred were only able to have one embryo retrieved and thus had poorer prognoses. However, again, we are modeling the decision for one woman, and as such, a one-embryo transfer would have, on average, at least one half the live birth rate of a two-embryo transfer. Thus we conservatively estimated the one-embryo live birth rate at one half that of the two-embryo live birth rate for women aged 38 years and older. For women aged younger than 38 years, live birth rates were reported in a CDC surveillance report from 2002 for a subset of women known to have more embryos available than transferred.13 Previous studies have shown that this subset represents women with more similar clinical prognosis.23 In this subset, for women aged younger than 35 years, the one-embryo live birth rate was 91.5% of the two-embryo rate, and for women aged 35–37 years, the one-embryo live birth rate was 62.7% of the two-embryo rate. Thus, for women aged younger than 35 years and for women aged 35–37 years we estimated the one-embryo live birth at 91.5% and 62.7%, respectively, of the two-embryo rate.
We used the 2002 CDC surveillance report to estimate the singleton, twin, and triplet rate by embryo transfer number. There were several aberrations in the data reflecting random variation due to the small number of women in the upper age groups that we smoothed to better estimate the outcomes in our model. For women aged 41–42 years, the multiple rate decreased for four- as compared with three-embryo transfers. We thus estimated the four-embryo multiple rate at halfway between the three and five-embryo rate. For women greater than 42, the three-embryo multiple rate decreased as compared with two-embryo transfers, so we estimated it at halfway between the two- and four-embryo rate. For women aged older than 42 years, the five-embryo multiple rate was less than the four-embryo rate; thus we estimated the five-embryo multiple rate as equal to that of the four-embryo rate.
We used a figure from a published study to estimate the percentage of women who would be able to transfer frozen embryos.14 In order for a woman to be able to undergo a frozen cycle, she must have retrieved at least twice the number of eggs as needed on the previous fresh transfer. Thus we used the number of women who would have 2, 4, 6, 8, and 10 embryos retrieved and capable of cleavage as an estimate of the number of women transferring 1, 2, 3, 4, and 5 embryos, respectively, who would be able to undergo a frozen cycle following a fresh cycle. We assumed the pregnancy rate for frozen transfers at 59% that of fresh transfers, as reported in the 2003 Assisted Reproductive Technology Report.
In terms of pregnancy outcomes, we modeled all preterm births, as well as births less than 32 weeks gestation, which may be more clinically meaningful. Estimates of preterm birth rates, by gestational number, and cerebral palsy rates, by preterm birth status, were taken from previous studies.15, 16
Parental costs included the full cost of IVF treatment. We estimated IVF costs by taking an average from five different IVF centers in Illinois, Texas, North Carolina, Massachusetts, and California (Advanced Fertility Associates, Santa Rosa, IVF treatment costs, unpublished. Personal communication, J. Ratcliffe, August 2005).17–20 We included the costs of the procedure, medication, cryopreservation for those women who are able to freeze embryos, and 56% of the sperm microinjection cost, because this is the percentage of women who use this technology, as reported in the CDC report. We also estimated indirect IVF costs by estimating the lost maternal productivity due to treatment time. We assumed that a complete cycle of IVF would take a total of four 8-hour workdays at the national average wage,24 and that a cycle from cryopreservation would take 2 days. This accounts for the both the retrieval and transfer and ultrasound monitoring visits, both prestimulation and poststimulation. In addition to IVF treatment costs, parental costs included $2,000 if there was a live birth. This assumes, at worst, parents would have an indemnity insurance product for obstetric and neonatal care with a 20% coinsurance and a $2,000 deductible.
Insurance costs included the short- and long-term costs of live births, minus the $2000 copay from the parents. For short-term costs, we used estimates of the total maternal and neonatal hospital costs for singleton, twin, and triplet births.4 Long-term costs included the direct lifetime medical costs and indirect morbidity and mortality costs (estimated by lost potential earnings) for those children who developed cerebral palsy.21
Societal costs included both the IVF treatment costs covered by the parents and the short and long-term costs of multiple births covered by the insurance companies. To reflect true societal costs, rather than charges, we multiplied IVF treatment costs by the average national ratio of costs to charges for inpatient care.22 The cost to charge ratio was used only for the societal perspective because both parents and insurance companies face charges rather than true costs. Although IVF treatment is an outpatient procedure, we assumed that due to the technological and procedural basis of the treatment that its cost to charge ratio was akin to inpatient care. All costs were projected to 2005 dollars using the medical component of the Consumer Price Index.25
We allowed a maximum of 5 cycles of IVF, but varied this number in sensitivity analysis. Total cost from each perspective was the primary outcome measure. As a secondary measure, we calculated costs per live birth. We also calculated clinical outcomes per 10,000 women: number of live births, singletons, twins, triplets, births less than 37 weeks of gestation, births less than 32 weeks of gestation, and the number of children who developed cerebral palsy.
We tested the robustness of our model by varying the maximal number of IVF cycles, performing univariable sensitivity analysis on total cost outcomes from a societal perspective, and running a Monte Carlo simulation. For the sensitivity analysis, we varied costs down to one half and up to 2 times baseline estimates. When varying each cost category, we maintained the baseline cost proportions. Our Monte Carlo simulation varied all probability estimates. We sampled probabilities during each recursive stage and ran the simulation a total number of times that reflected the total number of IVF cycles performed nationally for the given age subgroup.
The lowest total cost for every maternal age subgroup was to transfer one embryo (Table 2). However, the cost detriment to transferring more than one embryo varied by maternal age. For women aged younger than 35 years, it cost 80% more to transfer five compared with one embryo at a time ($39,212 total cost for five treatment cycles compared with $21,661). For women aged older than 42 years, it cost 13% more ($29,102 compared with $25,723).
The cost per live birth varied by maternal age. For women aged younger than 35 years, one-embryo transfers had the lowest cost per live birth ($24,216 for one embryo as compared with $41,907 for five embryo transfers). Whereas for women aged 35 years and older, the lowest cost per live birth was achieved with transfer of four or five embryos at a time. For example, for women aged older than 42 years, the lowest cost per live birth was achieved with five embryos ($70,176 for five-embryo transfers as compared with $235,343 for one-embryo transfers).
From a societal perspective, the incremental cost per live birth (the additional cost for achieving one more live birth by transferring one more embryo) also varied considerably by maternal age. This rate was calculated by dividing the additional cost of transferring two compared with one embryo or three compared with two, etc. by the additional live birth rate. For women aged younger than 35 years, it cost approximately $300,000 to achieve one more live birth by transferring two rather than one embryo and more than a million and one half dollars for one more live birth by transferring three rather than two embryos. However, for older women, there was less incremental cost. For women aged older than 42 years, it cost about $8,000 per additional live birth for two compared with one embryo and about $23,000 for three compared with two embryos.
From a parental perspective, transferring one embryo was the most expensive option. The least expensive option was between two and five embryos, depending on maternal age. Total costs were fairly similar for parents, regardless of embryo transfer number, but in terms of cost per live birth, the additional cost for transferring one embryo increased dramatically with increasing maternal age. For example, it cost only 8% more for women aged younger than 35 years to transfer one instead of five embryos, whereas it cost 400% more for women aged older than 42 years.
From the perspective of an insurer, which would bear the pregnancy costs minus copays and deductibles, the total costs were lowest for women of all ages to transfer one embryo at a time. Not surprisingly, because multiple gestations are more costly than singleton pregnancies; this was also true for cost per live birth.
One-embryo transfers markedly improved clinical outcomes. For example, two- compared with one-embryo transfers for women aged younger than 35 reduced preterm birth and cerebral palsy rates by 55% and 41%, respectively (Table 3). Younger women had a greater chance of experiencing negative outcomes when transferring a greater number of embryos, reflecting higher rates of multiple gestations. For example, if women aged younger than 35 years transferred two embryos at a time, 36% of their births would be multiples, 22% would be born before 37 weeks of gestation, 3% before 32 weeks of gestation, and 0.5% would result in at least one child developing cerebral palsy. In women aged older than 42 years, transferring two embryos would result in 7% multiples, 10% of the neonates born at less than 37 weeks of gestation, 1.4% less than 32 weeks of gestation, and 0.3% with cerebral palsy.
Sensitivity analysis on total societal costs showed our results to be robust (Table 4 and Table 5). Varying the maximal number of cycles from 1–7 did not affect our conclusions. Costs increased overall, but cost rankings remained the same. Univariable sensitivity analysis on cost estimates down to one half and up to two times baseline also showed our results to be robust. And in Monte Carlo simulation in which all probability estimates were varied simultaneously, one-embryo transfers were statistically significantly less expensive than two-embryo transfers for all age groups (P < .005).
Our study demonstrates that from the perspective of society as a whole, IVF costs are minimized when one embryo is transferred at a time. This is also true from the perspective of third-party payers. Parents, however, minimize their costs when between two and five embryos are transferred at a time, depending on maternal age. This demonstrates the importance of payor perspective in IVF treatment decisions. Parents face the costs of IVF treatment, and thus maximizing their live birth rate by transferring a greater number of embryos is the optimal economic decision. This economic decision is not a small one, often costing large portions of individuals' annual salary. Insurance companies face the cost of multiple births, and thus minimizing multiple birth rates by transferring the fewest number of embryos is the optimal economic decision. And from a societal point of view, which balances the treatment and IVF costs, the cost of multiple births predominates.
Although these findings may not be surprising, our study does underscore the importance of maternal age in terms of the magnitude of discrepancy among the various perspectives. For younger women, given their increased chance of getting pregnant and the increased risk of multiple gestations with more than one embryo transferred, it was relatively more expensive from a societal perspective for these women to transfer multiple embryos and lead to relatively worse clinical outcomes. However, for the older maternal age subgroups, the societal cost per live birth was actually less expensive with increased embryo transfer number. The incremental cost of an additional live birth by transferring two compared with one embryo for women aged older than 42 years was $8,000 as compared with $300,000 for women aged younger than 35 years. And the incremental cost per live birth of transferring three compared with two embryos was only $23,000 for women aged older than 42 years as compared with more that one and one half million dollars for women aged younger than 35 years.
How do our findings compare to clinical practice? Women aged younger than 35 years transfer an average of approximately three embryos per cycle, and women aged older than 40 years transfer an average of approximately four embryos per cycle.5 Thus, from a societal perspective, younger women, in particular, are transferring too many embryos. If a woman aged younger than 35 years transferred one at a time instead of the current national average of three, there would be societal savings of approximately $15,000 per woman treated. Further, her risk of multiple birth would fall from 42% to 2%, and she would have a four-fold reduction in preterm birth before 32 weeks of gestation and a two-fold reduction in the rate of cerebral palsy. However, from the parental point of view, this would cost an additional $1,100 due to need for more IVF cycles for some of the couples.
Our study is not without limitations. Although we corrected aberrations in the data, the probability estimates may still not perfectly represent the different rates of IVF success with variation in numbers of embryos transferred, because it is still based on cross-sectional data. Several trials have compared one to two embryo transfers, but have not examined one through five embryo transfers.26,27 Without a prospective, randomized trial of embryo number, such data cannot be generated clinically. Further, we varied these inputs through ranges that might be anticipated with such a trial and our results were robust. Of note, our input data were from 2003, because this was the most recent year available. Results regarding infertility treatment outcomes may vary slightly from these data.
There are also inherent limitations of modeling to represent a clinical picture, which may fail to capture a number of intangible factors. Our model incorporated simplifications to the clinical setting, both in terms of possible outcomes and the cost of these outcomes. We did not account for the possibility of fetal reduction, although because fetal reduction is generally not performed for twin gestations, and the majority of the costs were from excess twin gestations, this would probably not have a great effect on our conclusions. We also did not account for women who use donor eggs, as complete birth rate data were not provided for these women in the CDC data. Nonetheless, our findings were quite robust in the sensitivity analyses performed.
The disparity between the societal and parental perspectives, in particular, may represent one of the reasons why policy has not been particularly effective in reducing the risk of multiple birth associated with IVF. There are a number of ways that the societal, parental, and insurance perspective might be better aligned. First, insurance companies could provide better coverage for IVF treatment, but limit the numbers of embryos transferred. Studies have demonstrated that insurance coverage does seem to affect embryo transfer practice.28 Insurance companies may even save money, or least significantly blunt the cost of expanding coverage, if women transfer fewer embryos as a result of the increased coverage. Women aged younger than 35 years, for example, currently are transferring an average of three embryos, thus costing insurance companies an average of $50,000 for delivery costs. However, if insurance companies covered the cost of IVF treatment (thus assuming the IVF and delivery costs minus the delivery copay paid by the parents) and women only transferred one embryo at a time, insurance companies would spend an average of $43,000 and so would actually save about $7,000 per couple. This insurance coverage would also allow couples to make decisions about infertility treatment that represented their preferences as opposed to having financial pressures to maximize fertility at the cost of worse clinical outcomes.
We may also consider imposing regulations on the maximal number of embryos to transfer, as other countries have done.5 Alternatively, we might lessen the pressure on physicians to transfer a greater number of embryos. The public availability of clinic-specific live birth rates, the very data used to perform this study, puts pressure on IVF clinics to maintain a competitive live birth rate.29 Reporting practices have recently been challenged in the literature,30 and indeed the newest annual report by the CDC tries to address this concern partially by providing a new outcome statistic: singleton live birth rate. Finally, many physicians who practice in the realm of assisted reproductive technology believe that it is improvements in IVF itself that are going to lower the multiple birth rate. Improving IVF success rates may lessen the pressure to transfer more than one embryo, either by improving the implantation rate per embryo or by improving freezing technique so that more women are able to freeze embryos and don't have to start over after each single embryo transfer.
If policy seeks to reduce the multiple birth rate from IVF procedures, it must address the conflicting interests highlighted by our study. Especially for younger women, transferring more than one embryo at a time leads to dramatically higher societal costs and worse clinical outcomes, yet transferring one embryo at a time is the most expensive strategy for the parents. Future research needs to address these opposing financial incentives and examine policies that may facilitate resolution of this conflict. Such research should examine couples' preferences toward clinical outcomes, their decision-making process in the setting of infertility management, and effects on overall health outcomes and costs from increasing access to infertility treatment. From a policy standpoint, this is an issue that interested parties, parents, clinicians, and organizations that guide clinical care, such as the American College of Obstetrics and Gynecology or the American Society for Reproductive Medicine, should consider by encouraging improved decision making that incorporates long-term outcomes before the need for intervention from state or federal policy makers is required. As clinicians, we should encourage our infertile couples to strongly consider the possibility of improved outcomes from single embryo transfers and singleton pregnancies, despite the possibility of increased costs to them.
1. Centers for Disease Control and Prevention (CDC). From the Centers of Disease Control and Prevention. Use of assisted reproductive technology—United States, 1996 and 1998. MMWR Morb Mortal Wkly Rep 2002;51:97–101.
2. Hogue CJ. Successful assisted reproductive technology: the beauty of one. Obstet Gynecol 2002;100:1017–9.
3. Katz P, Nachtigall R, Showstack J. The economic impact of the assisted reproductive technologies. Nat Cell Biol 2002;4:S29–32.
4. Callahan TL, Hall JE, Ettner SL, Christiansen CL, Greene MF, Crowley WF Jr. The economic impact of multiple-gestation pregnancies and the contribution of assisted-reproduction techniques to their incidence. N Engl J Med 1994;331:244–9.
5. Jain T, Missmer SA, Hornstein MD. Trends in embryo-transfer practice and in outcomes of the use of assisted reproductive technology in the United States. N Engl J Med 2004;350: 1639–45.
6. Collins J. Cost-effectiveness of in vitro fertilization. Semin Reprod Med 2001;19:279–89.
7. De Sutter P, Gerris J, Dhont M. A health-economic decision-analytic model comparing double with single embryo transfer in IVF/ICSI. Hum Reprod 2002;17:2891–6.
8. Wolner-Hanssen P, Rydhstroem H. Cost-effectiveness analysis of in-vitro fertilization: estimated costs per successful pregnancy after transfer of one or two embryos. Hum Reprod 1998;13:88–94.
9. Lukassen HG, Schonbeck Y, Adang EM, Braat DD, Zielhuis GA, Kremer JA. Cost analysis of singleton versus twin pregnancies after in vitro fertilization. Fertil Steril 2004;81:1240–6.
10. Gerris J, De Sutter P, De Neubourg D, Van Royen E, Vander Elst J, Mangelschots K, Vercruyssen M, et al A real-life prospective health economic study of elective single embryo transfer versus two-embryo transfer in first IVF/ICSI cycles. Hum Reprod 2004;19:917–23.
11. Kalra SK, Milad MP, Klock SC, Grobman WA. Infertility patients and their partners: differences in the desire for twin gestations. Obstet Gynecol 2003;102:152–5.
12. Centers for Disease Control and Prevention, America Society for Reproductive Medicine. Assisted reproductive technology success rates, 2003. Available at: http://www.cdc.gov/ART/ART2003/nation.htm
. Retrieved June 7, 2006.
13. Wright VC, Schieve LA, Reynolds MA, Jeng G, Division of Reproductive Health, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention (CDC). Assisted reproductive technology surveillance—United States, 2002. MMWR Surveill Summ 2005;54:1–24.
14. Ziebe A, Loft A, Petersen JH, Andersen AG, Lindenberg S, Petersen K, et al. Embryo quality and developmental potential is compromised by age. Acta Obstet Gynecol Scand 2001;80: 169–74.
15. Multiple gestation pregnancy. ESHRE Capri Workshop Group. Hum Reprod 2000;15:1856–64.
16. Drummond PM, Colver AF. Analysis of gestational age of cerebral palsy in singleton births in north-east England 1970-94. Paediatr Perinat Epidemiol 2002;16:172–80.
21. Waitzman NJ, Romano PS, Scheffler RM. Estimates of the economic costs of birth defects. Inquiry 1994;31:188–205.
23. Schieve LA, Peterson HB, Meikle SF, Jeng G, Danel I, Burnett NM, et al. Live-birth rates and multiple-birth risk using in vitro fertilization. JAMA 1999;282:1832–8.
26. Thurin A, Hausken J, Hillensjo T, Jablonowska B, Pinborg A. Strandell A, et al. Elective single-embryo transfer versus double-embryo transfer in in vitro fertilization. N Engl J Med 2004;351:2392–402.
27. Pandian Z, Bhattacharya S, Ozturk O, Serour GI, Templeton A. Number of embryos for transfer following in-vitro fertilisation or intra-cytoplasmic sperm injection. The Cochrane Database Syst Rev 2004;4:CD003416.
28. Reynolds MA, Schieve LA, Jeng G, Peterson HB. Does insurance coverage decrease the risk for multiple births associated with assisted reproductive technology? Fertil Steril 2003;80:16–23.
29. Alper MM. In vitro fertilization outcomes: why doesn't anyone get it? Fertil Steril 2004;81:514–6.
© 2006 The American College of Obstetricians and Gynecologists
30. Schieve LA, Reynolds MA. What is the most relevant standard of success in assisted reproduction?: challenges in measuring and reporting success rates for assisted reproductive technology treatments: what is optimal? Hum Reprod 2004;19: 778–82.